use all pairs for fine-tuning, seq len 128

finetune.py

@@ -12,7 +12,7 @@ import gzip
 
 model_name = 'sentence-transformers/all-MiniLM-L6-v2'
 
-train_batch_size = 128
+train_batch_size = 100
 max_seq_length = 128
 num_epochs = 1
 warmup_steps = 1000
@@ -25,13 +25,12 @@ class ESCIDataset(Dataset):
         with gzip.open(input) as jsonfile:
             for line in jsonfile.readlines():
                 query = json.loads(line)
-                for i in range(1,10):
-                    if len(query['e']) > 0 and len(query['i']) > 0:
-                        p = random.choice(query['e'])
+                if len(query['e']) > 0 and len(query['i']) > 0:
+                    for p in query['e']:
                         positive = p['title']
-                        n = random.choice(query['i'])
-                        negative = n['title']
-                        self.queries.append(InputExample(texts=[query['query'], positive, negative]))
+                        for n in query['i']:
+                            negative = p['title']
+                            self.queries.append(InputExample(texts=[query['query'], positive, negative]))
 
     def __getitem__(self, item):
         return self.queries[item]

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8fd77ee0f10a3e91e0a3798f2b7b0ad1cf92538d062d089ea1fe83594dcdf5ab
+oid sha256:e4f93afe73f1a6e9b5b7d8ddae92a900d5c02208e215866f340aac1312d0d0e3
 size 90891565

pytorch_model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b9a477737434014d347f63a17bd90d092e3f0916e7b728e55ac887fd1b8b064c
+oid sha256:93063c8b319d7875dae261c2291d4a955f4025831b3bdced74fc6a5c7e2ee3da
 size 90984263